Automated Oil Palm Fruit

International Journal of Video & Image Processing and Network Security IJVIPNS-IJENS Vol: 11 No: 03
30
Automated Oil Palm Fruit Grading System using
Artificial Intelligence
Z. May1, M. H. Amaran2
Electrical and Electronic Engineering Department
Universiti Teknologi PETRONAS
31750, Tronoh, Perak, Malaysia
zazilah@petronas.com.my1, anep88@gmail.com2
Abstract - This project deals with the ripeness of oil palm fruit.
The current procedure in the palm oil mills is grading the oil
palm fruit manually using human graders. This method is
subjective and inconsistence because each graders has its own
techniques and may vary from each other’s. Hence, it affects
the quality and quantity of the oil that can be extracted. In this
project, a new model of automated grading system for oil palm
fruit is developed using the RGB color model and artificial
fuzzy logic. The purpose of this grading system is to distinguish
between the three different classes of oil palm fruit which are
underripe, ripe and overripe. The ripeness or color ripening
index was based on different color intensity. The grading
system uses a computer and a CCD camera to analyze and
interpret images correspondent to human eye and mind. The
computer program is developed for the image processing part
like the segmentation of colors, the calculation of the mean
color intensity based on RGB color model and the decision
making process using fuzzy logic to train the data and make
the classification for the oil palm fruit. The program developed
has been able to distinguish the three different classes of oil
palm fruit automatically with 86.67% of overall efficiency.
This project provides a very good technique to standardize the
oil palm fruit grading system over a large area and the
research will continue to normalized the system to be able to
use under different source of lighting.
Keyword- Classification, Grading, Color Model, Fuzzy
Logic
I. INTRODUCTION
Automated grading of agriculture products has been
getting special interest recently as the demand for higher
quality food products produced within a shorter period of
time has increased. Market grade of quality food products
are determined based on their multiple features: flavor,
texture and appearance. While flavor may be measured
using chemical compounds to determine the sweetness or
acidity, texture properties such as firmness and mouth feel
are difficult to measure. In automated fruit grading,
appearance (shape, color and size) is generally utilized to
classify the fruit’s grade
Color provides helpful information in estimating the
maturity and examining the freshness of fruits. Color is one
of the most significant criteria related to fruit identification
and fruit quality and it is a good indicator for ripeness. The
color of an object is determined by wavelength of light
reflected from its surface. In biological materials the light
varies widely as a function of wavelength. These spectral
variations provide a unique key to machine vision and
image analysis.
In Malaysia, researches in automated grading system
have becomes an interest for many researcher since it has a
high potential for a new approach in the future generation.
For the oil palm fruit research, there were several report has
been published in the past few years. The research has been
done using several techniques and the most recent one are
done by using RGB Digital Number[1] and by using NeuroFuzzy[2]. There are also other techniques used to evaluate
the grading of oil palm fruit[3-5]. Besides the oil palm fruit
research, there are also other fruit that has been studied and
developed in classifying the degree of fruit ripeness
automatically using image processing techniques. The
automated grading techniques for apple[6-7], banana[8],
orange[9] and other fruit[10-11] are some of the examples
of other fruit. Even though RGB color model, fuzzy logic
and neural network have been popularly employed for oil
palm grading, none of the work has tries to do it based on
the oil palm fruit. Most of them are focused on oil palm fruit
bunches.
II. METHODOLOGY
The grading system depends on the color extracted
from the image. Therefore, color features extraction plays
an important role in developing this grading system. The
flow of each processing level as well as the components
used are stated and briefly discussed in Fig. 1. The steps
taken for latter approach includes image acquisition, color
117203-3636 IJVIPNS-IJENS © June 2011 IJENS
IJENS
International Journal of Video & Image Processing and Network Security IJVIPNS-IJENS Vol: 11 No: 03
feature extraction and finally classification using fuzzy logic
algorithm.
Image
An image is captured
using Olympus E520.
Grading (Fuzzy
Logic)
The classification is
determined based on
extracted color
feature.
Background
Removal
The unnecessary
background is
removed and
converted to black.
Color Feature
Extraction
The features are
mean value of red,
green and blue
layers.
Fig. 1: The flowchart of grading process
A. Image Acquisition
The oil palm fruits are collected from a plantation. The
oil palm fruits are graded manually by human grader as
soon as he collects the fruits. A total of 75 images are taken
in the laboratory under controlled lighting using an
Olympus E-520 digital camera. They are converted into
JPEG format and resized to 640x480 pixel dimensions.
Even though the best lighting is using direct sunlight but we
are using the controlled lighting condition because this
system is planned to use in the building.
B. Background Removal
The region of interest is the fruit itself without the
background. The background is the noise to the image and it
has to be removed. The background subtraction method is
used to remove the white background. Fig. 2 shows the
result of performing the background removal.
underripe, ripe and overripe categories, we need to obtain a
range of mean value of red, green and blue layer for each
fruit. These ranges values are used as a reference and a
range input of fuzzy logic system. A total of 75 images (i.e
25 underripe, 25 ripe and 25 overripe) are used in
determining the range value of red, green and blue of each
category. The mean values of red, green and blue layers are
calculated using the following equations:
MeanR = R / No. of pixels
MeanG = G / No. of pixels
MeanB = B / No of pixels
MeanR
MeanG
MeanB
R
G
B
=
=
=
=
=
=
Mean value of Red layer
Mean value of Green layer
Mean value of Blue layer
Red pixel
Green pixel
Blue pixel
D. Grading
Fuzzy Logic is a chosen method for classifying the oil
palm fruit into underripe, ripe and overripe categories. This
technique is selected because it represents a good approach
when we want to interpret the decision making process of
human to the computer. Currently, the oil palm fruit grading
is done based on the experience of the human grader. This
method needs to be replaced with a new approach that is
able to standardize the grading process.
Fig. 3 shows the complete process of developing a
fuzzy inference system (FIS) for the grading process using
MATLAB[12].
Surface
Viewer
Fig. 2: Background removal of oil palm fruit
The color features of the oil palm fruit are analyzed
based on the RGB color model. To classify the fruit into
(1)
(2)
(3)
Where:
FIS Editor
C. Color Feature Extraction
31
Membership
Function Editor
Rule
Viewer
Rule Editor
Fig. 3: Fuzzy Inference System
The process consists of 3 main steps: defining the input
and output in Membership Function Editor, set the fuzzy
rules in Rule Editor and obtaining the output for each rule in
Rule and Surface Viewer.
117203-3636 IJVIPNS-IJENS © June 2011 IJENS
IJENS
Internatiional Journal of
o Video & Imaage Processingg and Network Security IJVIP
PNS-IJENS Vool: 11 No: 03
32
The gradinng system has three inputs (Red,
(
Green and
a
Blue) and one output
B
o
(Catego
ory). The membbership functioons
a builds usinng trapezoidal shapes since it gives the best
are
r
result
compareed to other shapes. A total of 17 ruules
s
statements
are created in order to classify the
t oil palm frruit
c
categories.
Fig 7: The Memberrship Function reprresentation of bluee input.
F 4: The Fuzzy Inference
Fig
I
System consists
c
of three innputs and one outpput.
Fig 8: The Membershipp Function represeentation of Categoory output.
Fig 5: The Membership Funcction representationn of red input.
Thhe fuzzy sets produce
p
17 rulees statements in
i order to
classifyy the oil palm fruit’s categorry. Examples of
o the rules
are illustrated as folloows:
T
TABLE
1. FUZZY
Y SETS RULES FO
OR CLASSIFICA
ATION
M
Functiion representation of green input.
Fig 6: The Membership
Rule 2
If Redd is low
And Green
G
is low
And Blue
B is medium
m
Then category is Unnderripe
Rule 12
If Redd is medium
And Green
G
is high
And Blue
B is medium
m
Then category is Rippe
117203-3636 IJVIP
PNS-IJENS © June 2011 IJENS
Rule 5
R is low
If Red
Andd Green is highh
Andd Blue is hedium
Then category is Ripe
R
Rule 16
R is high
If Red
Andd Green is highh
Andd Blue is mediuum
Then category is Overripe
O
IJENS
International Journal of Video & Image Processing and Network Security IJVIPNS-IJENS Vol: 11 No: 03
Fig. 4-8 below shows the rule viewers which consist of
the system’s input and output and the defuzzification result
column. The first to third column is the inputs which are
red, green and blue values while the last column is the
category column. The last row of the category column
shows the defuzzification result where category of oil palm
fruit is obtained.
33
III. RESULTS AND DISCUSSIONS
The grading system developed is managed to classify
the oil palm fruit into underripe, ripe and overripe
categories. The mean value of red, green and blue layers of
5 training data from each category is calculated and
tabulated in the graph below.
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Underripe
Ripe
Red
Overripe
Green
Blue
Fig. 10: RGB values for underripe, ripe and overripe of training image.
Fig. 9: Defuzzification result from the Rule Viewer
Based on the Defuzzification result from the Rule
Viewer from Fig. 9, the oil palm fruit fulfilled Rule 12
where red is medium, green is high and blue is medium. The
value of the category is calculated by using the centroid
method. Then the classification of oil palm fruit is
determined based on the crisp logic given in Table 2.
The RGB values are compared with each category in
order to determine the suitable range of RGB value for each
category. From Fig. 10, it is clearly seen that the three
difference categories have a different color distributions.
The overripe category has the highest red pixel, the ripe
category has an average red pixel while the underripe
category has the lower red pixel compared to overripe and
ripe categories. The condition also similar for green and
blue pixels but the difference between the three categories is
not as much as the red pixel.
The range values for the three different categories are
used in the fuzzy logic system as the reference. The range is
shown in the Table 3.
TABLE2. FUZZICICATION ALGORITHMS
TABLE 3. THE RANGE VALUE OF RGB OF THE TRAINING DATA
Defuzzification Output
(cat <= 65)
(cat >= 66) && (cat <= 155)
(cat >= 156)
Oil Palm Fruit
Category
Underripe
Ripe
Overripe
The value of cat is 100. Therefore, the oil palm used in this
experiment is classified as ripe category.
Red
Green
Blue
Category
Min
Max
Min
Max
Min
Max
Ripe
82.42
127.05
39.26
59.01
27.34
48.32
Underripe
52.53
81.95
19.85
33.22
15.26
26.88
Overripe
148.34
179.20
48.73
74.54
28.33
50.29
117203-3636 IJVIPNS-IJENS © June 2011 IJENS
IJENS
International Journal of Video & Image Processing and Network Security IJVIPNS-IJENS Vol: 11 No: 03
An example of misclassification of oil palm fruit is
shown in Fig. 11. Based on human graders, the fruit is
classified as underripe but the program classified it as ripe
category. This problem occurs because of the color of the
fruit lies in between underripe and ripe categories.
34
between two different categories and an additional feature
that incorporate differences between the categories such as
texture need to be added.
The results of the fuzzy logic system are evaluated
against the human graders to measure accuracy. It shows
that the automated oil palm fruit grading system using fuzzy
logic achieved 86.67% accuracy in overall categories. This
shows that the grading system using fuzzy logic have a high
potential of accuracy in grading the oil palm fruit.
IV. CONCLUSION
Fig. 11: The misclassification of oil palm fruit
The details result of oil palm fruit classification is
shown in Fig. 12. For red and green pixels, the values are
lies in between two sections where this situation can
misclassify the fruit.
The study has proven that by using fuzzy logic
algorithm, the accuracy of oil palm fruit grading is quite
high. A graphical user interface was developed for the user
to use the program easily. This technique begins with
loading the image to the program. The image is then
processed by image processing, calculated the mean RGB
value and graded using fuzzy logic. The oil palm is
classified based on RGB color. The results are shown on the
surface viewer from the Fuzzy Toolbox. The oil palm fruit
grading system can be improved by adding more training
image from various places and choose the best range of
RGB.
ACKNOWLEGDEMENT
We would like to thank Mr Shafie Shaari and Mr Faiz
Shafie for providing the oil palm fruit sample and for their
full cooperation and support throughout our research
project. The oil palm fruit are taken from the oil palm
plantation located at Batu 4, Jalan Bidor-Teluk Intan, Perak,
Malaysia.
REFERENCES
[1]
Meftah Salem M. Alfatni, “Oil Palm Fruit Bunch Grading System
Using Red, Green, and Blue Digital Number”, University Putra
Malaysia, Journal of Applied Sciences, Vol. 8, pp. 1444-1452, 2008.
[2]
Nursuriati Jamil, Azlinah Mohamed, Syazwani Abdullah,
“Automated Grading of Palm Oil Fresh Bunches (FFB) using
Neuro-Fuzzy
Technique”,
Universiti
Teknologi
MARA,
International Conference of Soft Computing and Pattern
Recognition, 2009.
Fig. 12: The misclassification result from Rule Viewer
[3]
In order to improve the system performance, the better
range for each color needs to be obtained. The fuzzy rules
need to be improved especially for the fruit that lies closely
W.I. Wan Ismail, M.Z. Bardaie, and A.M. Abdul Hamid, “Optical
properties for mechanical harvesting of oil palm FFB,” J. Oil Palm
Research , vol. 12, no. 2, pp. 38-45, Dec 2000.
[4]
A. Nureize , J. Watada, “Multi-criteria Fuzzy Regression Model
for Evaluating Oil Palm Grading”, Graduate School of
117203-3636 IJVIPNS-IJENS © June 2011 IJENS
IJENS
International Journal of Video & Image Processing and Network Security IJVIPNS-IJENS Vol: 11 No: 03
Information, Production and System, Waseda University, 2-7
Hibikino, Wakamatsu, Kitakyushu, 808-0135 Japan, 2009.
[5]
M.Z Abdullah, L.C Guan and B.M.N Mohd Azemi, “Stepwise
Discriminant Analysis For Colour Grading of Oil Palm Using
Machine Vision System”, University Science Malaysia, Institute of
Chemical Engineers, Trans ICheme, Vol. 79, Part C, 2001.
[6]
G. RENNICK, Y. ATTIKIOUZEL and A. ZAKNICH, “Machine
Grading and Blemish Detection in Apples”, University of Western
Australia, Fifth International Symposium on Signal Processing and
its Applications, ISSPA, 1999.
[7]
I. KAVDIR, D.E. GUYER, “Apple Grading Using Fuzzy Logic”,
Turk J Agric For 27, pp. 375-382, 2003.
[8]
Nur Badariah Ahmad Mustafa, Nurashikin Ahmad Fuad, Syed
Khaleel Ahmed, Aidil Azwin Zainul Abidin and Zainul Abidin Md
Sharif. “Image Processing of an Agriculture Produce: Determination
of Size and Ripeness of a Banana”, Universiti Tenaga Nasional,
2008.
[9]
Michael Recce and John Taylor, “High Speed Vision-Based Quality
Grading of Oranges”, Department of Anatomy and Developmental
Biology, University College London, London, UK, 1996.
[10]
Nur Badariah Ahmad Mustafa, Nurashikin Ahmad Fuad, Syed
Khaleel Ahmed, Aidil Azwin Zainul Abidin and Zainul Abidin Md
Sharif. “Agriculture Produce Sorting and Grading using Support
Vector Machines and Fuzzy Logic”, Universiti Tenaga Nasional,
IEEE International Conference on Signal and Image Processing
Applications, 2009.
[11]
Antonio Carlos Loureiro Lino, Juliana Sanches, Inacio Maria Dal
Fabbro, “Image Processing Techniques for Lemons and Tomatoes
Classification”, Bragantia, Campinas, v.67, n.3, p.785-789, 2008.
[12]
The Mathworks. Fuzzy Logic Toolbox User’s Guide, Version 2.2.13
(Release 2011a): The Mathworks, Inc, 2011.
35
Muhammad Hanif Amaran was born in
Terengganu, Malaysia in 1988. He obtained
his Foundation at Negeri Sembilan
Matriculation College. Presently he is a FYP
student under the guidance of Ms. Zazilah
May in University Teknologi PETRONAS,
Tronoh, Perak. B Eng (Hons) in Electrical
and Electronic Engineering. Majoring in
Power System and minor in Electrical
Machines.
BIOGRAPHIES
Zazilah May is a lecturer in the Department
of Electrical and Electronic Engineering,
Universiti Teknologi Petronas (UTP),
Malaysia. She graduated with BSc.(Hons)
in Mathematics (1998) from University of
Leicester, UK. She holds a Master degree
MSc. in Advanced Control (2000) from
UMIST, UK. She is currently pursuing her
PhD in Electrical and Engineering at UTP.
Her research areas are image processing,
video surveillance, computer vision and
artificial intelligence system. She can be
contacted at: zazilah@petronas.com.my
117203-3636 IJVIPNS-IJENS © June 2011 IJENS
IJENS